Open homanp opened 11 months ago
Hi @homanp it's a great idea. I'm writing it down for a future article, thanks! :)
@mlabonne I would heavily suggest SFT-ing this newly released model by Nexusflow, which surpasses GPT-4 according to their evals. There is a discussion thread on the page here, where they suggest that simple instruct tuning (ideally only on the completions) can go pretty far.
If you want, I could work on this when I have time over the next few weeks.
@vikram71198 Very interesting, please let me know if you do it!
ohh
Hey @mlabonne
Great work on this repo!
Would be amazing if you would want to add an example of fine-tuning for function calling. This would gain a lot of traction due to devs wanting OSS alternatives to GPT.